pdf2text

Extract text and diagrams from PDFs and images using Gemini AI.

Install

Executable (global):

npm install -g @nilsfischer/pdf2text

Library (project dependency):

npm install @nilsfischer/pdf2text

CLI Usage

Authenticate once (stores key encrypted in system keychain):

pdf2text auth set --api-key YOUR_GEMINI_API_KEY

Run extraction:

pdf2text input.pdf
pdf2text input.pdf --format latex --output ./results --split

Runtime key resolution order:

1. --api-key
2. GEMINI_API_KEY
3. Stored keychain key (pdf2text auth set)

If keychain support is unavailable, use GEMINI_API_KEY or --api-key.

Node Library API

Use full pipeline and filesystem APIs from the Node subpath:

import { runPipeline, parsePageRanges } from "@nilsfischer/pdf2text/node";

await runPipeline({
  input: "./slides.pdf",
  apiKey: process.env.GEMINI_API_KEY!,
  format: "markdown",
  outputDir: "./output",
  split: false,
  noDiagrams: false,
  pages: parsePageRanges("1-3"),
  scale: 2,
  concurrency: 5,
  verbose: false,
});

Web Library API

Use browser-safe extraction APIs from the web subpath:

import { extractFromImageBytes } from "@nilsfischer/pdf2text/web";

const imageBytes = new Uint8Array(await file.arrayBuffer());
const result = await extractFromImageBytes(imageBytes, {
  apiKey: "YOUR_GEMINI_API_KEY",
  format: "markdown",
  noDiagrams: false,
});

console.log(result.text);
console.log(result.diagrams);

Shared API

Shared helpers available from @nilsfischer/pdf2text:

• parsePageRanges
• getFormatInstructions
• getFormatExample
• replacePlaceholders
• buildGeminiPrompt

Options

| Flag | Alias | Default | Description | |------|-------|---------|-------------| | -f, --format | | markdown | Output format: markdown, latex, or text | | -o, --output | | ./output | Output directory | | -s, --split | | false | Split output into per-page files | | --diagrams | | true | Enable diagram detection (use --no-diagrams to disable) | | -p, --pages | | all | Pages to process, e.g. 1-5,8,12-14 | | --scale | | 2 | PDF render scale factor (2 ≈ 144 DPI) | | -c, --concurrency | | 5 | Max parallel Gemini API calls | | --api-key | | keychain/env | Gemini API key (flag overrides keychain/env) | | -v, --verbose | | false | Verbose logging output |

Local Development

bun install
npm run build
node dist/cli.js --help

Publish

npm publish